Is an AI retention paradox hiding in equity vesting?

By Nigel Watson

8 Apr 2026 6 min read

I'm starting to wonder if companies are making the same AI mistake.

Just as AI is increasing the value of senior people who can direct work, challenge outputs and resolve uncertainty, traditional equity vesting potentially makes those same people easier to lose.

I suppose that's not really an AI problem. It's probably more of an incentives design problem.

For years, the default employee equity model (certainly in the UK) was simple enough: grant equity, vest over three or four years and assume ownership will do the retention work. In a more stable, slower paced world, that looked good enough. In an AI-shaped one, maybe it isn't.

The premium is moving from production to orchestration

Having gone through the looking glass on this one (so you don't have to), a lot of AI commentary still seems to start from the wrong premise. It assumes the main effect of AI is to reduce the need for people. In some cases it will. But in many businesses, the more immediate effect is different. Yes, execution will become cheaper and faster, but judgement and experience will become more valuable.

Which in turn changes where the talent premium sits.

The people who matter most are increasingly those who can break work into the right tasks, supply the right context, test outputs, challenge confident‑sounding nonsense, apply commercial judgement and decide what can safely go out into the world. In other words, not just people who can produce work, but people who can orchestrate it.

That tends to favour experienced engineers, product leaders, data specialists, infrastructure architects, and senior operators - the people who ensure AI‑generated work is still good, not merely fast.

If that thesis is right, it should be visible. Boards can map where role leverage is rising fastest in an AI‑enabled operating model, align that with first meaningful vesting dates, and ask a simple question: does retention pressure spike at that point?

If it does, the company may already have a problem hiding in plain sight.

Vesting is not always the exit event. But it is often the moment risk rises

People do not usually walk out the door the day options vest. In private companies, vested options may remain illiquid. In listed companies, holding periods, dealing restrictions or tax timing can defer value. So vesting does not automatically mean cash in hand.

My point is more subtle than that.

Vesting is often the point at which an employee begins to reassess the ‘work relationship'. The original “golden handcuffs” logic weakens. The award feels earned. External opportunities start to feel real rather than hypothetical. The question gradually changes from “Can I leave?” to “What comes next?”

If the company does not have a clear answer, it can discover that what it assumed was a retention tool is, in fact, a retention cliff.

That is especially true where vesting aligns with any of the following:

meaningful value already delivered;
no visible follow-on / refresh awards;
frustration over liquidity or exercise cost;
a stronger external market for experienced operators; or
a business entering a phase where seasoned judgement matters more than raw output.

In other words, the technical issue is not vesting alone. It is the interaction between vesting, liquidity, psychology and marketability - and that interaction is where many equity strategies potentially fail.

AI may make the cliff steeper

To be clear, this is not a claim that AI has invented retention cliffs. Vesting‑related reassessment has always existed. The difference is that AI changes who creates the most value and how quickly that value becomes externally visible. Where judgement, orchestration and risk‑management skills now scale output, the gap between internal reward momentum and external opportunity can open faster and at precisely the point traditional equity design starts to decelerate.

There are signs of this happening in practice. In one fast‑growing client of ours, a cohort of senior engineers hired and granted equity within the same 12‑month window hit their first meaningful vest just as AI tooling materially increased their market worth. None left on vesting. But over the following six months, external approaches intensified and attrition among that group rose sharply, despite strong engagement scores and competitive pay.

What boards and remco's should actually review

If this risk is real, the answer is not simply “grant more equity”. It is to test whether the reward architecture still reflects where value is genuinely being created.

In practice, that means asking some hard questions.

1. Who is the genuinely scarce population?

Not all attrition matters equally. Which roles now carry the highest coordination and judgement premium? Who makes AI‑enabled scale safer, faster or more commercially useful? If that value increasingly sits with a more experienced layer, the reward design may need to recognise that explicitly rather than relying on broad‑based notions of alignment.

2. Where is the first meaningful vest date?

Plans are often assessed in aggregate, but risk concentrates around the first vesting point that feels economically real to the participant. That date should be treated as a potential attrition trigger, not merely an administrative milestone.

3. How concentrated is the grant cycle?

Where large numbers of critical employees were hired and granted at the same time, the company may have created a cohort risk. A single vesting pattern can quickly become a single retention failure point.

4. What is the refresh policy?

Too many businesses treat refresh grants as episodic fixes rather than core architecture. If there is no credible second‑cycle economics visible before the first cycle matures, vesting can start to look like the end of the story rather than the middle.

5. What is the liquidity policy?

Particularly in private companies, appeals to “long‑term alignment” may not be enough. Is there a credible path to occasional realisation? A structured secondary, partial cash‑out, buyback mechanism, or at least a transparent house view on likely liquidity windows? Illiquidity can retain, but it can also corrode belief and, for some types of companies, belief is really important. Read more here.

6. Is the instrument still the right one?

Options - essentially a leveraged bet on share price growth - are not always the right tool. Depending on stage and objectives, RSUs, restricted shares, growth shares, phantom awards or cash‑linked long‑term incentives may do a better retention job. Instrument choice should follow the commercial problem being solved, not legacy habit.

The real risk?

In a fast‑moving, AI‑dominated environment, the most important question for boards is increasingly simple: does the company’s reward design still reflect where value is actually being created?

If AI is elevating the importance of people who can orchestrate, verify and translate AI‑generated work, then incentive plans built on a generic, pre‑AI retention template may now be misaligned. They may still look orthodox. They may still benchmark comfortably. But they may be protecting yesterday’s scarcity, not tomorrow’s capability.

That seems to be a real risk to me.

Put simply, reward for past contribution is not the same thing as incentive for future contribution. A plan that does the first but neglects the second is only doing half the job and, in an AI‑accelerated business, that half may matter more than ever.

At Burges Salmon, we are seeing more clients ask not just whether their incentive plans are competitive, but whether they still fit the operating model of the business they now have. That is the right question. In an AI-shaped economy, effective incentive design is not just about equity delivery. It is about identifying where judgment sits, where retention cliffs exist and whether vesting, refresh, instrument choice and liquidity policy still support the capabilities the business cannot easily afford to lose.